REEXAMINATION OF ‘RESTRAINED VS.

UNRESTRAINED’

Kevin LaMalva 1, Luke Bisby 2, John Gales 3, Thomas Gernay 4, Elie Hantouche 5, Cliff Jones 6, Ali Morovat
7
, Robert Solomon 8 & Jose Torero 9

Simpson Gumpertz & Heger Inc., USA, corresponding author 1

University of Edinburgh, UK 2
York University, CAN 3
Johns Hopkins University, USA 4
American University of Beirut, LBN 5
Jensen Hughes, USA 6
University of Texas at Austin, USA 7
National Fire Protection Association, USA 8
University of Maryland, USA 9

Keywords

Structural fire protection, restrained, unrestrained, thermal restraint, structural fire engineering,
prescriptive method, standard fire resistance design.

Summary

For furnace testing of fire-resistant floor and roof assemblies in the United States, the ASTM E
119 standard (and similarly the UL 263 standard) permits two classifications for boundary
conditions: “restrained” and “unrestrained.” When incorporating tested assemblies into an actual
structural system, the designer, oftentimes a fire protection or structural engineer, must judge
whether a “restrained” or “unrestrained” classification is appropriate for the application. It is
critical that this assumption be carefully considered and understood, as many qualified listings
permit a lesser thickness of applied fire protection for steel structures (or less concrete cover for
concrete structures) to achieve a certain fire resistance rating if a “restrained” classification is
confirmed, as compared with an “unrestrained” classification. The emerging standardization of
structural fire engineering practice in the United States will disrupt century-long norms in the
manner to which structural behavior in fire is addressed. For instance, the current edition of the
ASCE/SEI 7 standard will greatly impact how designers consider restraint. Accordingly, this paper
serves as an exposé of the “restrained vs unrestrained” paradigm in terms of its paradoxical nature
and its controversial impact on the industry. More importantly, potential solutions toward industry
rectification are provided for the first time in a contemporary study of this paradigm.

KEYWORDS
prescriptive method, restrained, standard fire resistance design, structural fire engineering,
structural fire protection, thermal restraint, unrestrained

1 | INTRODUCTION

For furnace testing of fire resistant assemblies, ASTM E 119 (and similarly UL 263) permits two
boundary conditions: “restrained” and “unrestrained.” When incorporating tested systems into an
actual structural system, the designer, oftentimes a fire protection or structural engineer, must
judge whether a “restrained” or “unrestrained” classification is appropriate for the application. It
is critical that this assumption be carefully considered and understood, as many qualified listings
permit a lesser thickness of a given applied fire protection to achieve a certain fire resistance rating
if a “restrained” classification is confirmed, as compared to an “unrestrained” classification.

The emerging standardization of structural fire engineering practice in the U.S. will disrupt the
century-long norms in the manner in which structural behavior in fire is addressed. Notably, the
current edition of the ASCE/SEI 7 standard (Minimum Design Loads and Associated Criteria for
Buildings and Other Structures) has commenced a new industry-consensus standard of care for
structural fire protection, which will greatly impact how designers consider Fire Resistance and in
particular structural behavior such as restraint. Specifically, ASCE/SEI 7-16 explicitly addresses
what designers have known for decades: there is no correlation between assembly structural
performance in a furnace test and in-situ structural system performance under fire exposure. This
fact is absolutely irrefutable. In light of this, the “restrained vs. unrestrained” paradigm may be
regarded as paradoxical, and designers are rightfully concerned about the liability associated with
making uncertain judgments in this regard. Hence, proper rectification of this paradigm is
necessary as proposed herein.

Many listings in the UL Fire Resistance Directoryi permit reduced fire protection thicknesses to
achieve fire resistance ratings if the designer can demonstrate that the assembly will be “restrained”
when it is constructed as part of an actual structural system. Otherwise, the assembly must be
considered “unrestrained.” Architects routinely task structural engineers and/or fire protection
engineers to judge what classification should be assigned, often with pressure to select the less
restrictive ‘restrained’ classification. This paper will examine the paradoxical nature of this
paradigm. Furthermore, this paper offers potential solutions toward industry rectification on this
controversial issue.

The Need for Rectification

A structural solution is deemed to adequately perform in the event of a fire if it adheres to

prescriptive code requirements (referred to as standard fire resistance design), or less commonly
if performance is demonstrated by conducting structural analyses at elevated temperatures
(referred to as structural fire engineering). These two design methods share little to no
qualification metrics that can be cross-plotted for comparison purposes. For standard fire resistance
design, the primary qualification metric is ‘fire resistance,’ expressed in hours. This metric is an
artifact of standard fire testing, that requires thermal exposure to a standardized temperature
history and a failure criterion. The standardized temperature history does not equate to any realistic
fire and the failure criterion is generally expressed in thermal, not structural, terms. Even if
structural terms are used to define a failure condition, this is generally not representative of system
structural behavior. Therefore, ‘fire resistance’ has no correlation to actual structural system
performance at elevated temperatures. Conversely, structural limit state qualification metrics are
required for structural fire engineering. In essence, standard fire resistance design is an empirical
indexing method and structural fire engineering is structural engineering for a specific loading
case. Hence, cross-plotting their primary qualification metrics is not only inappropriate, but
specifically prohibited by new industry standards. Due to these irrefutable facts, the paradigm
known as ‘restrained vs. unrestrained’ within standard fire resistance design can be regarded
oversimplified, and the true implications of its form may not have been entirely contemplated
when first introduced. Consequently, this paradigm has perpetually perplexed designers and
building authorities alike.

The current “restrained vs. unrestrained” paradigm requires compatibility between standard fire
resistance design and structural engineering in order to be viable. However, there are many
incompatible aspects as examined herein. Standard fire resistance design is entirely based on
furnace testing of isolated structural components/assemblies, and relies almost exclusively on
insulation for structural fire protection. Accordingly, the intrinsic fire endurance (or weakness) of
a given structural system is not quantified or explicitly contemplated. Also, the anticipated in-situ
fire conditions are not quantified or explicitly contemplated. Rather, a single intense heating
exposure (i.e., ASTM E119ii/UL 263iii curve) is used to comparatively test isolated small-scale
mock-ups against specified acceptance criteria that are meant to generalize the robustness of the
protection scheme to severe fire exposure. The primary intent of this approach is to reduce the
heating of individual structural components, with the intent of mitigating the risk of structural
system failure during an uncontrolled fire. The primary mechanisms leading to structural failure
require an adequate definition of temperatures and temperature gradients, thus the link between
the intensity of heating and structural behavior is not direct or simple5.

As an alternative to standard fire resistance design, structural fire engineering explicitly evaluates
the demand and capacity of structural systems under in-situ fire conditions, in a similar manner as
other design loads are treated in structural engineering practice. Within this framework, the
demand on a structural system under fire exposure can be reduced by means of rationally-allocated
structural insulation (i.e., membrane protection or direct application methods) or control of fuel
loads. Also, the capacity of a structural system to endure fire effects can be increased by means of
reinforcement strengthening/placement/detailing, slab thickening, connection enhancements,
member sizing, geometric layout modifications, and/or other measures to enhance structural
robustness with respect to explicit performance objectives. As opposed to the binary “restrained
vs. unrestrained” classification system in standard fire resistance design, the analysis of structural
system response required for this alternative approach inherently considers the degree and
condition of restraint.

Origin of Restraint Classifications

The origin of the “restrained vs unrestrained” paradigm can be traced to the time period spanning
from 1966 to 1971, although the industry seems to have been aware of restraint effects since at
least 1899.7

The concept of restraint classifications originated when it was formally observed and documented
that the behavior of structural members and assemblies during a standard fire test can be
significantly influenced by the end restraint provided by the testing furnace frame. In this context,
restraint was identified as the combination of reactive moments and forces generated by the dead
and live loads plus the forces generated by temperature variations in structural members during
the fire test.8 Restraint was perceived to be generated at the perimeter of the test assembly as the
result of the method of framing and tolerance of fit in the test furnace.9

To address the impact of end restraint on the results of standard fire tests of floor and roof
assemblies, several revisions were proposed to the governing ASTM committee in 1964, 1965,
1967, 1968, and 1969 for adoption in the ASTM E119 standard.10 These proposals stemmed from
the observation that the results from fire tests on structural assemblies tested under identical
configurations, protection schemes, and temperature conditions varied considerably from test to
test.11 Development of a varying degree of applied restraint against thermal expansion in standard
fire tests was assumed to be the mainreason for such variations12 .

Additionally, during this time, there were debates about the feasibility of simulating realistic
restraint within a test furnace.13 The 1969 proposed revision received the greatest support from
the committee members and was first adopted in the 1971 edition of ASTM E119.14 Apparently,
results and observations from fire tests performed at Ohio State University with sponsorship from
the American Iron and Steel Institute (AISI) played a significant role inthe success of the 1969
proposal. The Ohio State tests were conducted on protected steel beams with both “restrained” and
“unrestrained” end conditions under the standard time-temperature exposure.15

The experimental program at Ohio State University involved fire tests of 12 beam-slab assemblies
with different degrees of composite action (non-composite to fully composite) subjected to varying
levels of end restraint. The range of end restraint varied from zero to full end-fixity. Each beam-
slab assembly consisted of a 4-in. concrete slab (36 in. wide) cast over a 22-gage steel deck and
supported by a 12 WF27 ASTM A36 steel beam. The steel beam was protected with spray-applied
insulation. The beam-slab assemblies had a length of approximately 15 ft (“restrained” assemblies
were about ó in. shorter). Sliding- and fixed-hinge bearings were used to model the end supports
of the “unrestrained” assemblies. For “restrained” assemblies, the end conditions were modeled
using the standard AISC B-Series bolted clip angle connections and fully welded end plate
connections.9

Prior to each fire test, point loads were applied at outer quarter points of the beam span for each
beam-slab assembly. The magnitude of the applied loads was calculated to impose design
allowable stresses to the beam-slab assemblies.Several observations were made from the fire tests
conducted at Ohio State University that later served as the basis for the 1969 proposed revisions
to the ASTM E119 standard. The most pertinent observation was on the effect of end restraint on
the performance of beam-slab assemblies during the fire tests. On average, a 25% increase in the
level of fire resistance was observed for the “restrained” beam-slab assemblies, as compared with
“unrestrained” assemblies. This increase was attributed to the negative bending moments induced
at the beam-ends in the “restrained” beam-slab assemblies. Specifically, the generated negative
bending moments at the beam-ends resulted in smaller effective positive bending moments at the
mid-span of each assembly.15 The 1969 proposed revisions to the ASTM E119 standard are
summarized in Table 1. As indicated in Table 1 (left column), restrained classifications were
proposed to be defined for two classes of structural systems: individual beams and floor/roof
assemblies. Also shown in Table 1 (right column) are proposals on how to obtain the. level of fire
resistance for each assembly type. Essentially, the 1969 proposed revisions recommend that beam
and floor/roof assembly tests be performed under restrained conditions and unrestrained
classifications be derived on the basis of application of temperature limitations during restrained
tests. Revealingly, this same proposal acknowledged that correlating restraint conditions of a
furnace test tothat present in actual building construction is a daunting task.10

Code Adoption and Application of Restraint Classifications

Restraint classification first appeared in ASTM E119-71iv. At this time, there were three
predominant model building codes used regionally in the United States: (1) the Basic National
Building Code published by the Building Officials and Code Administrators, International
(BOCA), in the northeast, (2) the Standard Building Code published by the Southern Building
Code Congress, International, in the southeast, and (3) the Uniform Building Code (UBC)
published by the International Conference of Building Officials, in the west. Investigating the
history of these building codes reveals that the 1970 edition of the BOCAv, the 1970 edition of the
UBCvi, and the 1969 edition of the Southern Standard Building Codevii each adopt and reference
ASTM E119 for fire-resistant construction. Beyond these externally pointing references, the
adopted buildings codes at the time of the revision to ASTM E119 in 1971 do not offer guidance
on restraint classification, thus abdicating the classification indistinctly. Although an interpretation
by BOCA in 1993 in essence clarified the need to apply the ASTM E119 criteria to the end
conditions (i.e., consideration of furnace boundaries), the codes remained silent on what
specifically constitutes “restrained” or “unrestrained”viii in in-situ construction. Complacency with
this conundrum at the time may have been based on a level of comfort with pushing the onus of
resolving the specific issues onto designers. This became clear as further code commentary was
developed on this paradigm.

In its first appearance in ASTM E119-71, guidance on restraint classification was provided in an
appendix to the standard, §A4. This restraint classification guidance remains in an appendix to this
date, now §X3, and the language remains largely unchanged. To date, ASTM E119 provides no
precise way to define “restrained” and “unrestrained” conditions for a given fire-rated assembly.
However, the standard offers the following limited guidance on restraint classifications. In the
context of testing, §X3.3 notes that “a restrained condition is one in which expansion and rotation
at the ends and supports of a load carrying test specimen resulting from the effects of the fire are
resisted by forces external to the test specimen.” Later §X3.5 describes restraint in actual building
construction as “floor and roof assemblies and individual beams in buildings are considered
restrained when the surroundings or supporting structure is capable of resisting substantial thermal
expansion and rotation throughout the range of anticipated elevated temperatures caused by a fire.”
In contrast, “unrestrained” conditions are essentially defined through a diagnosis of exclusion.
Namely, all conditions that are not considered to be “restrained” are considered to be
“unrestrained.”

Through the close of the 20th century, the three regional model building codes were abandoned
and the International Building Code (IBC)ix was developed as an amalgamation of the three
regional codes. Like it’s legacy model building code predecessors, the IBC also references ASTM
E119 (and UL 263) for fire-resistance ratings of building elements, components, or assemblies in
§703.2, and provides guidance on restraint classification in §703.2.3 where it states that “fire-
resistance-rated assemblies tested under ASTM E119 or UL 263 shall not be considered to be
restrained unless evidence satisfactory to the building official is furnished by the registered design
professional showing that the construction qualifies for a restrained classification in accordance
with ASTM E119 or UL 263.” Thus, the IBC, like its legacy model codes, abdicates responsibility
for defining restraint for in-situ conditions. Also, further confusion is added by including the
requirement that the registered design professional – typically a professional engineer – be
responsible for providing sufficient evidence to the building official – the authority having
jurisdiction (AHJ) – to qualify a material, assembly, or system as “restrained.”

To help fill the void of guidance/prescription in building codes, Underwriter Laboratories (UL)
publishes a companion guide to UL 263 that provides additional information for fire resistance
ratings, UL 263 (BXUV) Guide Information for Fire Resistance Ratings. In §III.15, the guide
specifies that restrained conditions in 14-ft by 17-ft fire test frames built from composite steel and
concrete offer an approximate flexural stiffness of 850,000 kip-in and 700,000 kip-in along their
short and long sides, respectively, which is significantly higher than that provided by most structural
systems. This stiffness remains essentially constant throughout a typical fire test because the test
frame is insulated from the fire environment. Also, in a somewhat circular reference, guide states
the following:

It is up to the designer and code authority to determine if an assembly is being used in a

restrained or unrestrained application, as required by the building code being enforced.

While helpful, the information above does not provide the explicit guidance necessary to
adequately judge restraint conditions in the context of the prescriptive method, no matter the
skill/competence of the designer.

Furnace Testing vs. In-Situ Restraint

Failure Criteria
At the heart of the divergence between in-situ and furnace testing restraint is the definition of
failure. UL 263/ASTM E 119 establish different criteria for failure being the main criterion is a
load bearing criterion that states that: “The test specimen shall have sustained the applied load
during its classification period without developing unexposed surface conditions which will ignite
cotton waste.” This criterion is identical for restrained and un-restrained. ASTM E 119 will further
specify temperature thresholds that shall not be exceeded and these temperature thresholds are
identical between restrained and unrestrained systems.
For example ASTM E 1191 indicates for un-restrained systems: “For test specimens employing
steel structural members (beams, open-web steel joists, etc.) spaced more than 4 ft (1.2 m) on
centers, the temperature of the steel structural members shall not have exceeded 1300°F (704°C)
at any location during the classification period nor shall the average temperature recorded by four
thermocouples at any section have exceeded 1100°F (593°C) during the classification period.”
Repeating the exact language and thresholds for restrained systems only changing the period from
the “classification period” to the “first hour.” ASTM E 119 further adds that: “For restrained
assembly classifications greater than 1 h, these temperature criteria shall apply for a period of one
half the classification period of the floor or roof construction or 1 h, whichever is the greater.” The
change associated to the period can only be justified in terms of an expected beneficial effect of
restraint on structural performance which implies, not only a quantification of the benefit in terms
of tolerable temperatures but also that structural behavior within a furnace is representative of the
structural system behavior. No clear evidence supporting either requirement is currently available.

A final difference occurs during the interpretation of “The test specimen shall have sustained the
applied load.” In a furnace test, the test is stopped when an arbitrary threshold is exceeded in
deflection or rate of deflection. At that time, it is deemed that failure is reached, as the structural
component can no longer sustain the load. The adopted thresholds are arbitrary, as they have no
particular meaning outside of the furnace testing context, and the values selected for these
thresholds may influence the fire ratingx, but it is necessary to adopt a limit to define the “end
point” of the test. There is no point in continuing the test because, on the one hand, that would
endanger the integrity of the equipment and, on the other hand, no redistribution of load is possible
since the test is conducted on an isolated structural component. In an actual structure, on the
contrary, the end conditions of a structural component affect the structural failure mode. While a
simply supported beam with no restraint will exhibit a failure mode that is qualitatively similar to
that of a furnace test, a restrained component part of a structural assembly may develop second
order load bearing mechanisms when undergoing large displacements under fire exposure that
dramatically change the failure mode and that render irrelevant the adoption of deflection threshold
criteria. For instance, a restrained beam may develop catenary action (i.e., a transition from
predominantly bending forces to tensile forces) provided the end conditions support the resulting
tensile forces. This type of behavior can be highly beneficial to the fire performance. Robust
structures can be designed to take advantage of catenary action, tensile membrane action, and other
second order and load redistribution mechanisms that build on structural system behavior to
enhance performance under fire (and under other extreme structural loading as well). The
mobilization of these potentially beneficial mechanisms depends on the restraint conditions in in-
situ structures; however, these effects are by no means captured by the “restrained” furnace tests.

Adequacy of the Furnace Restrain

In a standard furnace test, an assembly is considered “restrained” if it bears directly against the
edges of the furnace at the outset of the test. As mentioned, the flexural stiffness of furnace framing
is very high, and this stiffness remains constant throughout the fire test because the test frame is
insulated from the fire environment. Further, if the assembly is made of two components (e.g. in a
composite floor beam and concrete slab assembly), both components would be in contact with the
edges and restrained equally by the furnace framing during a “restrained” furnace test. In contrast,
an assembly in a standard test is considered “unrestrained” when it is free to thermally expand
without contacting the furnace edges.
The prescriptive definition of “restrained” fails extraordinarily to capture the complexity and
variety of actual restraint conditions observed in-situ during a fire. To provide tentative guidance
about mapping this prescriptive definition to actual building construction, UL 263 specifies that
"floor-ceiling and roof-ceiling assemblies and individual beams in buildings should be considered
restrained when the surrounding or supporting structure is capable of resisting substantial thermal
expansion throughout the range of anticipated elevated temperatures." However, there is an
acknowledged need for engineering judgment in assessing what constitutes “substantial thermal
expansion”. With in-situ construction, the restraint conditions of structural components during a
fire may be affected by many factors. These conditions may provide any degree of restraint
intermediate between “resisting substantial thermal expansion” and allowing free thermal
expansion and rotation, and further, they may significantly change during the fire event. Finally,
the restraint conditions eventually may enhance or decrease the resulting structural performance.

The effects of restraint provided on a structural component by an actually-constructed assembly

depend, amongst others, on the position of the support. For instance, for reinforced or prestressed
concrete slabs or beams, a horizontal axial restraint force may have a positive effect on the fire
behavior, provided that the line of thrust is below the resultant of the compressive stress block.
Axial restraint in horizontal concrete members can also enable compressive membrane action (or
arching effect) to develop, which enhances the load-carrying capacity of these members, but the
efficiency of this effect again depends on the location of the restraint on the edges of the memberxi.
In a composite floor beam and concrete slab assembly, unlike in furnace testing where both
components would be restrained equally, the beam and slab may experience varying degrees of
restraint. This can result in differential longitudinal movement under fire exposure, particularly if
the structural components are not acting compositely.

As mentioned, the degree of restraint provided on a structural component by an actually-

constructed assembly may vary during the fire event. Nonlinearities in the response of the
structural system at the ends of the component can result from the behavior of materials at elevated
temperatures. For instance, yielding of connections in a moment resisting frame modifies the
rotational restraint at the ends of the beam. Similarly, geometric nonlinearities (i.e., instabilities,
large displacements) can change the stiffness of an assembly. Considering a floor beam undergoing
thermal expansion as part of a structural system, this beam may impose lateral loading on the girder
and column support points. Depending on the characteristics of the support points, the thermal
expansion may be resisted or the thermally-induced lateral loading may exceed either the beam or
support capacity. Indeed, the possibility that the restraint thermal forces may jeopardize the
surrounding structure must be taken into account. For instance, in the 2010 fire in the Tour d’Ivoire
in Montreux, Switzerland, thermal elongation of concrete slabs subjected to the action of two
burning cars led to the collapse in shear of a column that was several meters away from the fire
sourcexii. Finally, the other components in the assembly (girders, column support points,
surrounding beams and floor) may also be subjected to heating due to the fire, in which case their
stiffness is affected by temperature and subject to thermal expansion, which may generate complex
transient force interactions at the boundaries of the different components.

The simplification of the fire exposure into a monotonously increasing gas temperature evolution
as used for standard furnace testing represents another limitation in the way that restraint is
considered by the prescriptive approach. It is fundamental that actual building fires consist of a
heating phase followed by a cooling phase. During the heating phase, the fire will increase the
temperature of the structural element in a manner that temperature gradients form within these
elements. The increase in bulk temperature corresponds to thermal expansion while the gradients
lead to curvature and overall contraction of the element27. The performance of the structural
element when restrained is therefore defined by the integrated effect of bulk heating and
temperature gradients. Given that a furnace follows a predefined temperature history, it cannot
reproduce fire related temperature gradients. Cooling generates a reversal of thermal strains (shift
from thermal expansion to contraction) that can affect connections between componentsxiii and
dramatically redistribute forces in a structural systemxiv. Supports may need to resist thermal
contraction as tensile forces build up in the structural system and through the connections. Under
realistic temperature histories, it is clear that the restraint forces at the end of a structural
component completely differ from those experienced during a furnace test where the component
expands against the edges of a frame of constant stiffness.

Experimental Evidence
Aside from first principles of structural mechanics, a number of published furnace test results
demonstrate the inadequacy of the prescriptive method’s treatment of restraint conditions. AISC
and AISI funded furnace testing of steel floor assemblies, which demonstrated that restraint from
the furnace frame provided no fire resistance benefit in the specific cases testedxv. This testing
resulted in modifications to a specific UL listing (D982). NIST performed furnace testing of steel
trusses typical of the WTC floor constructionxvi. They found that an unrestrained assembly
achieved a higher fire resistance rating when compared to an equivalent restrained assembly. The
NIST tests also raised concern about whether furnace tests are “scalable” to larger floor systems.
More recently, NIST performed localized fire tests on steel beams with differing end restraints and
found that the presence of restraint, provided through double-angle connections, decreased the fire
resistance of the beam when compared to the unrestrained configurationxvii. These test results
demonstrate that the effect of restraint varies among different structural systems and restraint
conditions, and cannot be easily simplified in practice, especially in a binary fashion.
Proper evaluation of restraint effects requires consideration of the in-situ structural system, which
is not represented in standard furnace testing used for the prescriptive method. The corollary is
that testing of isolated structural components with simplified and constant boundary conditions
does not properly evaluate in-situ restraint effects. In-situ restraint is complex to evaluate, and it
plays a paramount role in structural fire response. The importance of restrained conditions on the
fire performance of building structures has been established by full-scale fire tests such as the
Cardington testxviiixixxx as well as by analysis of real failures such as the NIST investigation on the
WTC 7 collapsexxi. These studies, and many subsequent analyses, showed that system behavior
differs dramatically from isolated member behavior, and that thermal expansion effects govern
much of the response of structures in fire. It is the objective of structural fire engineering to
quantify these effects, amongst others, by the application of scientific and engineering methods.
Currently, advanced analysis based on first principles and numerical methods is the only option to
fully address the behavior of structural system under fire. It allows capturing the complex
interactions between structural components, including the effects of thermal
expansion/contraction, the nonlinearities and large displacement effects, as well as capturing the
effect of realistic fire exposures including the different stages of fire development. As a result,
advanced analysis can account for the transient effect of restraint for different structural systems
and restraint conditions. The last two decades have seen major advances in the development of
advanced analysis software for structural fire engineering, with these tools being instrumental in
supporting structural fire engineering.

Examination of the Ongoing Justification

Realization that the prescriptive method does not –and theoretically cannot– position designers to
make coherent/accurate judgments with regard to the “restrained vs. unrestrained” paradigm is not
new to the building design community. In fact, both government and industry have repeatedly
advocated for more comprehensive guidance/prescription on the “restrained vs. unrestrained”
paradigm. However, these recommendations have yet to be acted on. In the final report on the
World Trade Center Investigationxxii, NIST recommended (Group 2, Enhanced Fire Endurance of
Structures, Recommendation 5) that improvements to the methods for fire resistance testing should
be evaluated including the “effect of restraining thermal expansion (end-restraint conditions) on
test results.” A related recommendation was put forth in 2011 as part of a workshop that was
organized to establish an agenda for the National Fire Research Laboratoryxxiii. A group
recommendation from the workshop dealing with steel structures provides a rather blunt view of
the issue as follows:

Research should be undertaken to properly characterize the catenary and membrane

action behavior that currently is poorly captured by the restrained vs. unrestrained
argument. The group believes that the [paradigm] is meaningless when one looks at the
system behavior.

A recent article in a structural engineering magazine echoes the above concernxxiv. Also, during a
recent workshop on the “restrained vs. unrestrained” paradigmxxv, the following salient points were
expressed:
• From the code authority perspective, furnace testing standards only provide
examples/guidance for restraint classifications, so the designer is ultimately responsible for
such judgments.
• From the fire testing standards authority perspective, the boundary conditions of the test
furnace should be referenced when making a judgment on restraint within the prescriptive
method.
• From the academic perspective, restraint conditions of a furnace test differ from those of
the in-situ structural system.
• From the designer perspective, although consideration of restraint is a common task in the
industry, designers are concerned about the liability associated with making uncertain
judgments.

Historically and currently, restraint classification in U.S. building codes and their referenced
standards remains essentially undefined. Limited guidance is provided; however, it is insufficient
to ensure uniform application across a range of materials, components, assemblies and building
construction types. As a result, the need for further guidance remains in 2019. Amidst the
confusion, industry groups such as the American Iron and Steel Institute (AISI), and American
Institute of Steel Construction (AISC), advocate that all typical steel construction should be rated
as restrainedxxvi based on recommendations provided in limited researchxxvii. Such justification is
primarily based on the observance of beneficial compressive membrane action during furnace
testing, which acts to limit member deflections and achieve a higher fire resistance rating.
However, it is known that any compressive membrane action achieved by in-situ structural systems
tends to break down during the early stages of heating, and floors are more likely to act as tensile
membranes thereafter, assuming that adequate support/continuity is provided. Hence, it is clear
need that building codes and fire testing standards need to take an active role in explicitly defining
in-situ “restrained” conditions and regulating the interpretation to ensure uniformity in application
across all materials, constructing types, and geographic regions. In essence, a line needs to be
“drawn in sand” for it is paradoxical for a designer to do so based on first principles of structural
mechanics.

Contrary Industry Advances

ASCE/SEI 7-16xxviii permits designers to utilize structural fire engineering as an alternative to the
code-default prescriptive method. Specifically, Section 1.3.7 states that structural fire protection
shall be provided per prescriptive requirements of the applicable building code, or by employing
a performance-based approach in accordance with the new Appendix E section per building
authority approval. This standard reiterates/reaffirms/crystallizes the irrefutable fact that there is
no correlation between assembly performance in a furnace test and in-situ structural system
performance under fire exposure. Accordingly, the standard prohibits designers from
intermingling aspects of the prescriptive method with structural fire engineering. For instance,
Section CE.2 states that standard fire testing does “not provide the information needed to predict
the actual performance of a structural system during structural design fires.”

ASCE/SEI 7-16 also addresses thermal restraint more directly/explicitly than preceding U.S.
codes/standards to date. Notably, Sections E.2 and CE.6 state that the “level of restraint depends
on the adjacent framing and connection details,” which are excluded from standard fire testing,
and that “thermal restraint may dominate the behavior of framing systems, particularly floor
systems, with degradation of stiffness and strength a secondary factor.” Further, it is stated that
thermal restraint “may generate forces sufficient to cause yielding or fracture, depending on the
temperature reached and the degree of restraint provided by the surrounding structural system to
the thermally-induced actions.” Hence, this standard clarifies that a restrained condition can be
worse than a relatively unrestrained condition with all else equal, and that thermal restraint
conditions encompass infinitely more than two distinct scenarios (i.e., “restrained” and
“unrestrained”).

Effectively, Appendix E brings structural engineers into the fold of structural fire protection design
when alternatives to the prescriptive method are sought by project stakeholders. Notably, the
inclusion of Appendix E in ASCE/SEI 7 marks the first time that fire effects are considered as an
explicit design load condition in a U.S. structural engineering standardxxix. Overall, ASCE/SEI
guidance should validate structural engineers whom wish to engage and lead in the field of
structural fire protection. Also, building officials now have tools to comprehensively evaluate
structural fire protection variancesxxx. It is the authors’ opinion that such guidance will also
influence building officials’ interpretation and enforcement of “restrained” within standard fire
resistance design.

As a supplement to Appendix E, ASCE/SEI has recently released a new Manual of Practice No.
138 (Structural Fire Engineering)xxxi. Similar to ASCE/SEI 7-16, MOP-138 prohibits designers
from intermingling aspects of the prescriptive method with structural fire engineering. Notably,
Section 7.2.1 states that “designers should analyze the level of restraint from adjacent structural
framing that would resist the thermal expansion of a heated assembly or subsystem and not
extrapolate standard fire test results to evaluate the restraint condition of a structural system.”
Further, Section 2.2.1 states that “binary restraint classification may be difficult (or even
paradoxical) for a designer to judge with relation to in-situ conditions considering the incompatible
aspects of standard fire resistance design and structural fire engineering.” Even more directly,
Section 7.2.1 states that “designers should not use standard fire test results for evaluating the
restraint condition” of an assembly as part of a structural system. If design industry consensus is
sought on the “restrained vs. unrestrained” paradigm, the standardized language above should
leave nothing open to interpretation. However, building codes are still held to the “restrained” and
“unrestrained” qualifications inherent in the UL Directory and other similar sources for fire
resistant assemblies.

Proposed Rectification

There is an unavoidable conflict between the boundary conditions and general limitations of
standard furnace testing to represent realistic mechanical behavior, and the understanding of how
structural systems actually perform under fire exposure. ASCE/SEI 7-16 and ASCE/SEI MOP-
138 highlight this conflict, for they both express that “restrained” is a condition that does not exist
outside of a test furnacexxxii. Hence, the pressure that designers routinely experience from
stakeholders to make a “restrained” judgment is not to apply the scientific method and their innate
engineering knowledge, but rather to endorse a fallacy within a paradox with the backing of their
credentials/stamp. This is truly unfair to designers, especially those that fully contemplate the
relevant aspects discussed herein. For this reason, the “restrained vs. unrestrained” paradigm
continues to frustrate designers, and many remain conflicted between their ethical duty as a
registered design professional to base their judgments on code prescriptions or engineering
factuality, and stakeholder pressures to not deviate from the “norm,” which actually varies
geographically. The current paradigm sets designers up for failure by its definition, and does not
even incrementally raise the fire safety of buildings as compared to an environment in which it
does not exist.

In the authors’ opinion, the current prescriptive provision permitting reduced fire protection
thicknesses for a “restrained” assembly is in need of proper rectification. At the practical level, the
current paradigm’s reliance on a designer’s judgment is highly problematic. At the scientific level,
the prescriptive provision is challenged by experimental evidence and analytical reasoning that
must not be ignored for convenience. Fundamentally, the current paradigm directly conflicts with
the philosophy of new standardization which explicitly prohibits the selective adoption of aspects
of the prescriptive method and structural fire engineering. The prescriptive method is not
performance-based in any way, and so the “restrained vs. unrestrained” paradigm should not be
construed as such.

Similar to most other aspects of the prescriptive approach, an industry-consensus prescription on

the matter within the building code and/or the ASTM E119 standard (and similarly UL 263) would
relieve designers of an unrealistic obligation of assessing what/when constitutes an in-situ
“restrained” condition. However, it should be noted such a prescription would be entirely arbitrary
and not based upon any postulation of actual structural system performance under fire exposure.
As an alternative, the entire “restrained vs. unrestrained” paradigm could be abolished. In this case,
the industry would need to obtain a consensus on whether fire resistance listings dependent upon
this paradigm would default to current “restrained” (typically less insulation thickness) or
“unrestrained.” This would require the next edition of the UL Directory to be reformatted in this
respect.

In summary, either an explicit and arbitrary prescription on what “restrained” is should be enacted,
or the entire paradigm be should deleted entirely from applicable codes and standards with proper
adjustment to fire resistance directories. Paralysis on this matter is not an option, for the calls for
reform will continue as industry advancement accelerates in the U.S.xxxiii.

i
UL Fire Resistance Directory, Underwriters Laboratories, Northbrook, IL, 2015
ii
ASTM E119 (2016). E119-16a Standard Test Methods for Fire Tests of Building Construction and Materials, ASTM
International, West Conshohocken, PA.
iii
ANSI/UL 263 (2011) Standard for Fire Tests of Building Construction and Materials, Underwriters Laboratories,
Northbrook, IL.
iv
ASTM E119 (1971). E119-71 Standard Test Methods of Fire Tests of Building Construction and Materials, ASTM
International, West Conshohocken, PA.
v
The BOCA Basic Building Code, Building Officials & Code Administrators International, Inc., Chicago, IL, 1970.
vi
Uniform Building Code (UBC), International Conference of Building Officials, Whittier, CA, 1970.
vii
Southern Standard Building Code, Southern Building Code Congress, Birmingham, AL, 1969.
viii
The BOCA Basic Building Code, Building Officials & Code Administrators International, Inc., Chicago, IL, 1993
ix
International Building Code (IBC), International Code Council (ICC), Washington, DC, 2018.
x
Dumont, F., Wellens, E., Gernay, T., Franssen, J.M. (2016). Loadbearing capacity criteria in fire resistance
testing. Materials and Structures, 49(11), 4565-4581.
xi
Moss, P. J., Dhakal, R. P., Wang, G., & Buchanan, A. H. (2008). The fire behaviour of multi-bay, two-way reinforced
concrete slabs. Engineering Structures, 30(12), 3566-3573.
xii
Burnier, O. (2011). Reconstitution de l’incendie de deux voitures dans le parking de la Tour d’Ivoire à Montreux,
le 9 décembre 2010, Travail de diplôme, heig-vd ed., Yverdon-les-Bains.
xiii
Garlock, M. E., Selamet, S. (2010). Modeling and behavior of steel plate connections subject to various fire
scenarios. Journal of Structural Engineering, 136(7), 897-906.
xiv
Gernay, T., Gamba, A. (2018). Progressive collapse triggered by fire induced column loss: Detrimental effect of
thermal forces. Engineering Structures, 172, 483-496.
xv
C.J. Carter, F. Alfawakhiri, “Restrained or Unrestrained?” Modern Steel Construction, Chicago, IL, September
2013
xvi
Gross, J., Hervey, F., Izydorek, M., Mammoser, J., Treadway, J. (2005). Fire Resistance Tests of Floor Truss
Systems. Federal Building and Fire Safety Investigation of the World Trade Center Disaster, NIST NCSTAR.
xvii
Ramesh, S., Seif, M., & Choe, L. (2017, November). Localized fire tests on steel beams with different end restraints.
In The 9th International Symposium on Steel Structures.
xviii
Lennon, T., Moore, D. B., Bailey, C. (1999). The behaviour of full-scale steel-framed buildings subjected to
compartment fires. The Structural Engineer, 77(8), 15-21.
xix
Gillie, M., Usmani, A. S., Rotter, J. M. (2001). A structural analysis of the first Cardington test. Journal of
Constructional Steel Research, 57(6), 581-601.
xx
Gillie, M., Usmani, A. S., Rotter, J. M. (2002). A structural analysis of the Cardington British steel corner test.
Journal of Constructional Steel Research, 58(4), 427-442.
xxi
Gann, Richard G. (2008). Final Report on the Collapse of World Trade Center Building 7, Federal Building and
Fire Safety Investigation of the World Trade Center Disaster (NIST NCSTAR 1A). No. National Construction Safety
Team Act Reports (NIST NCSTAR).
xxii
NIST NCSTAR 1, “Final Report of the National Construction Safety Team on the Collapse of the World Trade
Center Towers”, Recommendation 5, September 2005.
xxiii
Almand, K.H, “Structural Fire Resistance Experimental Research – Priority Needs of U.S. Industry”, Prepared for
the Engineering Laboratory National Institute of Standards and Technology.
xxiv
LaMalva, K.; McAllister, T.; Bisby, L., “Restrained vs. Unrestrained: Within the Context of New Industry
Guidance,” STRUCTURE Magazine, September 2018
xxv
LaMalva, K., “Reexamination of Restrained vs. Unrestrained,” FPE Extra, Issue #20, Society of Fire Protection
Engineers, August 2017.
xxvi
Carter, Charles J.; Alfawakhiri, Farid, “Restrained or Unrestrained?” Modern Steel Construction, September 2013.
xxvii
Gewain, Richard G.; Troup, Emile W., “Restrained Fire Resistance Ratings in Structural Steel Buidlings,”
Engineering Journal, Second Quarter, 2001.
xxviii
ASCE/SEI 7: Minimum Design Loads and Associated Criteria for Buildings and Other Structures, American
Society of Civil Engineers: Structural Engineering Institute, Reston, VA, 2016
xxix
Post, N. “Guidance for Structural Fire Engineering Making Its Debut,” Engineering New Record, New York, NY,
February 2017
xxx
Post, N. “Fanning the Flames for Structural Fire Engineering,” Engineering News Record, New York, NY, January
2018.
xxxi
ASCE/SEI Manual of Practice No. 138: Structural Fire Engineering, American Society of Civil Engineers:
Structural Engineering Institute, Reston, VA, 2018
xxxii
LaMalva, K.J.; McAllister, T.; Bisby, L., “Restrained vs. Unrestrained: Within the Context of New Industry
Guidance,” STRUCTURE Magazine, September 2018.
xxxiii
CPF Grant #04-18. Advancing Performance-Based Structural Fire Engineering Design in the U.S. through
Exemplar Procedural Guidance, Charles Pankow Foundation, Vancouver, WA, 2018 <
https://www.pankowfoundation.org/projects/advancing-performance-based-structural-fire-engineering-design-in-
the-u-s-through-exemplar-procedural-guidance/>